VibeSkills — Intelligent Skill Orchestrator
Automatically routes local Skills and orchestrates workflows, so complex tasks need no manual planning.
- Source repo
- foryourhealth111-pixel/Vibe-Skills
- Stars
- ★ 3.7k
- Last updated
- today
- License
- Apache-2.0
- Primary language
- Python
- FA score
- 46/100 · Major gaps
At a glance
- How it runs
- Works with
- Portable with changesCodex · Claude Code
- Cost
- Free software; you pay for model usage
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- Data scientists: running a reproducible machine-learning experiment that requires data audit, statistical review, and production of figures and a scientific report.
- Not a fit if
- Users on platforms other than Claude Code or Codex who won't adapt it
- Users wanting zero configuration with no Skill folders or state files
- Source review
- 46/100 · Major gaps 1 safety controls not found
What does this agent do, and when should you use it?
VibeSkills is a general-purpose Skill that sits on top of AI assistants to automate the discovery and orchestration of local Skills. It scans configured Skill folders, decomposes complex tasks into modular work units, and assigns the right Skills to each part, avoiding scheduling conflicts between the harness framework and domain Skills. It follows design patterns from frameworks like Superpower and GSD. VibeSkills offers two workflow levels: L for manageable multi-step tasks, and XL for larger tasks with parallelizable parts. It records plans, assignments, and execution progress in JSON files such as agent_skill_organization, module-work-plan.json, module-execution.json, and delivery-acceptance-report.json, enabling continuity and final acceptance checks. It can be invoked via $vibe or /vibe within supporting AI applications. The project is Apache-2.0 licensed and includes a real-world machine-learning experiment case with detailed documentation.
VibeSkills scans local Skill folders (including additional roots listed in ~/.vibeskills/skill-roots.json or workspace .vibeskills/skill-roots.json), extracts names and summaries, and shortlists candidate Skills for each task module. During planning, it writes agent_skill_organization and module-work-plan.json to define responsibilities, expected outputs, and checks. During execution, it records actual assignments in module_assignments and progress in module-execution.json. For code tasks, it can employ test-driven development: failing test, change, re-test. After execution, it runs verification checks (e.g., required-files, dataset-contract, baseline-results) and compares results against the plan, producing a delivery-acceptance-report.json or .md. Outputs include data audits, figures, reports, and slide decks. It supports multiple Skills including third-party ones but never auto-invokes all of them.
- Data scientists: running a reproducible machine-learning experiment that requires data audit, statistical review, and production of figures and a scientific report.
- Project managers: decomposing complex projects into modular workflows with automated assignment of local Skills, reducing manual planning burden.
- Developers: using test-driven development within a TDD loop for coding tasks, with automatic verification.
- Teams: sharing and reusing team or third-party Skills without constraints from a single framework.
- Content creators: generating bilingual summaries, slide decks, and ensuring consistency across deliverables.
How do you install or deploy this agent?
Installation steps: 1. Clone the repository: git clone https://github.com/foryourhealth111-pixel/Vibe-Skills.git. 2. Follow the installation documentation (recommend using PowerShell to run the install script; exact commands are in the docs). After installation, run pwsh ./check.ps1 to verify runtime state. Pre-configure Skill folders in ~/.vibeskills/skill-roots.json or workspace .vibeskills/skill-roots.json.
How do you use this agent?
Usage steps: 1. In any AI application supporting local Skills, invoke VibeSkills via its Skills entry using $vibe or /vibe (or the syntax provided). 2. It discovers Skills, plans the task, and recommends a workflow level (L or XL). 3. Approve the plan; it executes and records progress. 4. Review the delivery acceptance report to confirm all checks pass. Detailed examples are in the quick-start guide and the ML experiment case.
What are this agent's strengths and limitations?
- Automated Skill orchestration: selects and assigns Skills per module, eliminating manual per-step Skill choices.
- Reduced planning overhead: modularizes complex tasks, so users don't need to explain Skill usage to the AI in detail.
- Parallel execution support: XL workflow can run up to two non-conflicting parts concurrently.
- Detailed execution records: JSON files document plan, assignments, and results, facilitating session recovery and review.
- Learning curve: users must understand workflow levels, Skill root configuration, and the planning state machine.
- Configuration dependencies: requires pre-configured Skill folders and Skills with non-conflicting names and readable SKILL.md.
- Context overhead: with large Skill libraries or complex tasks, token usage increases, though mitigated by local indexing and on-demand reads.
- Platform lock-in: currently documented for frameworks like Claude Code and Codex; other platforms may need adaptation.
How does this agent compare with similar options?
Similar to frameworks like Superpower and GSD, but VibeSkills emphasizes automated Skill orchestration and modular decomposition, maximizing reuse of local Skill assets.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| VibeSkills — Intelligent Skill Orchestrator This agent | 46 · Major gaps | Agent plugin / skillFree + model costs | ★ 3.7k | today | Python | Codex · Claude Code |
| AIWG | Insufficient evidence | CLIFree + model costs | ★ 224 | today | TypeScript | Codex · Claude Code |
| Autoprompt Skill | 69 · Some gaps | CLIFree + model costs | ★ 1.2k | 15d ago | JavaScript | Codex · Claude Code |
| OpenTeams | 56 · Major gaps | Desktop appFree + model costs | ★ 583 | 5d ago | TypeScript | Codex · Claude Code |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows: README describes requirement confirmation, plan approval, execution recording, and final check, indicating user confirmation; but no specific implementation details for least privilege, sensitive data handling, dependency security, external effects control, rollback, or source attribution. Deductions: lack of clear evidence for permissions, data flow, dependency security, external side effects, rollback, and source attribution.
Evidence shows: README and CI workflow (vco-gates.yml) provide coherent process description and verification steps, but no specific examples of failure messages or dependency availability guarantees. Deductions: insufficient evidence for failure messages and dependency availability.
Evidence shows: README describes target audience (users of AI applications with local Skills), applicable scenarios (complex tasks), capability boundaries (does not automatically call all Skills), and environment fit (supports local folder configuration). But trigger precision (e.g., specific syntax for $vibe or /vibe) is not detailed. Deductions: insufficient evidence for trigger precision.
Evidence shows: README provides information architecture (documentation index), install notes (docs/install), examples (ML case), known limitations (token overhead), and license (Apache-2.0). But evidence for naming stability, maintenance responsibility, and version changelog is insufficient. Deductions: insufficient evidence for naming stability, maintenance responsibility, and version changelog.
Evidence shows: README describes output usability (deliverable checks), marginal value (reduces planning burden), and cost-benefit (token overhead control). But specific cost-benefit data (e.g., token usage) is not provided. Deductions: insufficient evidence for cost-benefit.
Evidence shows: README and CI workflow provide verification steps and case, but do not clearly separate facts from inferences, and cross-source verification is insufficient. Deductions: insufficient evidence for fact-inference separation and cross-source corroboration.
- Not found in source: sensitive-data handlingUse dedicated, low-privilege, revocable API keys — never production credentials — and keep secrets out of logs.
- Publisher identity is unverified; treat as unknown and do not rely on brand trust.
- Static review cannot verify actual execution, test results, or security; use with caution.
- The case and CI results in README were not independently verified in this review and may be overstated.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
v4.1.1 Vibe Skills 4.1.1
Release notes excerpt
Vibe Skills 4.1.1 Highlights This patch release collects the runtime-correctness and installer-safety fixes merged since v4.1.0, including PRs #313, #315, #317, #319, #321, #323, and #325, plus the merged-main CI repair in #328. - Share read-only Skill source authority across Python discovery and canonical admission, including configured workspace/user roots, enabled state, metadata validation, controller exclusion, duplicate precedence, and resolved path containment. …View on GitHub ↗ -
v4.1.0 Vibe Skills 4.1.0
Release notes excerpt
Vibe Skills 4.1.0 - Original release date: 2026-08-29 - Asset refresh date: 2026-08-31 - Release commit: ddcaa2affca93c1efe026d008b6b93709e5fb7e2 Highlights Vibe Skills 4.1.0 brings the strongest Skill-organization and delivery patterns from the SkillsBench work into the existing production workflow. - Carries user-confirmed Skill inputs, outputs, planning hints, and verification hints from Skill Cards into WorkPlan and ModuleAssignments. - Adds explicit module dependencies with stable topological ordering, keeping execution and delivery aligned with the approved plan. …View on GitHub ↗ -
v4.0.0 Vibe Skills 4.0.0
Release notes excerpt
Highlights Vibe Skills 4.0.0 is the major Agent execution-handoff release. - Replaces retired native and delegated execution lanes with Agent-owned module handoff. - Requires one canonical module-execution.json return through vibe re-entry. - Keeps delivery acceptance closed when required modules are failed, blocked, or incomplete. - Removes the legacy vibe-what-do-i-want, vibe-how-do-we-do, vibe-do-it, and vibe-upgrade public entry names. Use vibe. - Hardens exact task preservation and avoids injecting coding or TDD requirements into non-code work. …View on GitHub ↗